Texture-tracking camera method exposes mould risk in mycelium insulation

Mould grew on every single specimen of a mycelium-based insulation material tested under wet, humid conditions, and a new image-analysis technique captured the colonisation as it happened, often before human eyes could see it. The study, published in Biotechnology for Sustainable Materials, tested mycelium-based composites made from hemp-shiv and Ganoderma curtisii mycelium and offers a low-cost, objective way to monitor mould on bio-based building materials whose naturally uneven surfaces confound conventional visual scoring.

Mycelium-based composites are bio-fabricated by growing fungal mycelium, the network of filamentous hyphae that forms the vegetative body of a fungus, on lignocellulosic substrates such as hemp shiv. As the fungus colonises and partly digests the substrate, it secretes a natural adhesive that binds the particles into a solid material, which is then dried in an oven to kill the fungus and leave a non-living bio-based product. These composites have attracted attention as low-carbon insulation alternatives to expanded polystyrene and mineral wool, materials that can carry high embodied carbon and depend on finite resources. Recent research has reported promising thermal conductivity, fire safety and life-cycle results, and several companies are developing mycelium insulation products, although most are not yet commercially available.

Durability, however, remains a critical gap. Bio-based materials are rich in cellulose, hemicellulose, lignin and proteins, and can therefore serve as a nutrient source for mould fungi, which need moisture, warmth and nutrients to grow. Mould on building materials poses health risks and carries social and economic costs, and a significant proportion of homes experience damp or mould, with impacts particularly severe for vulnerable occupants. The prevailing recommendation for prevention is environmental control, lowering humidity and temperature and improving ventilation, but this becomes harder as buildings are sealed to reduce heat transfer and during weather extremes. Standard mould susceptibility tests rely on subjective rating scales, typically scoring coverage from 0 to 5, which is prone to observer bias and poorly suited to tracking mould progression over time, especially on heterogeneous materials where natural colour variation can mask early growth.

The research team, led by Joni Wildman with Valeria Cascione, Daniel Henk and Andrew Shea, produced twenty mycelium-based composite discs in Petri dishes using pre-chopped hemp-shiv combined with plain white flour at a 10:1 weight ratio, sterilised by autoclave and inoculated with Ganoderma curtisii grain spawn. After eight days of incubation at 23 degrees Celsius, the specimens were dried at 50 degrees Celsius for 48 hours, reducing their mass by an average of 80.3 percent. Fifteen specimens went into mould susceptibility testing and five were reserved for thermal measurements. The finished material had a density of 77.6 kilograms per cubic metre and a thermal conductivity of 0.040 watts per metre-kelvin, figures comparable to both bio-based and conventional insulation materials and consistent with earlier studies showing that mycelium composites can effectively reduce heat transfer.

The mould test was adapted from the BS EN 17886:2023 standard for assessing the susceptibility of thermal insulation products to mould growth, deliberately modified to simulate more extreme conditions such as leaks or severe weather events. Ten specimens and three control media plates were inoculated with a mixed spore suspension totalling one million spores per millilitre, drawn from six fungal strains obtained from the Muséum national d’Histoire naturelle: Aspergillus niger, Trichoderma viride, Talaromyces pinophilus, Chaetomium globosum, Paecilomyces variotii and Aspergillus versicolor. After the suspension dried, each specimen was wetted with one millilitre of sterile water, and five un-inoculated control specimens were wetted identically. Everything was then incubated at 90 percent relative humidity and 25 degrees Celsius for ten days under continuous illumination.

Throughout the incubation, a 48-megapixel camera mounted 45 centimetres above the samples captured images every 60 seconds. The researchers analysed frames at hourly intervals using ImageJ, computing Grey-Level Co-occurrence Matrix, or GLCM, texture metrics for each specimen. The GLCM is a statistical tool that quantifies how often pairs of adjacent pixel intensities occur together in an image, and its contrast metric measures the degree of local variation in grey levels. As mould develops, it disrupts the uniform surface appearance of a specimen, increasing pixel-to-pixel heterogeneity, so rising contrast provides a quantitative signature of fungal colonisation. Unlike area-based segmentation methods, which depend on mould colonies presenting clear pigment contrast against the background, GLCM contrast detects changes in surface texture that can shift before visible colony formation, making it particularly suited to materials with inherently variable surfaces. Contrast values were normalised to the first image of each sequence, and contrast correlated strongly with the other GLCM features extracted, with Pearson coefficients of -0.926 against angular second moment, -0.908 against correlation and 0.958 against entropy.

All ten inoculated specimens showed visible mould growth, confirming the material’s susceptibility under moisture stress, while the five un-inoculated controls showed no visible growth and their contrast values remained stable throughout. To validate the image metric against human perception, the team ran a survey in which 20 postgraduate students and academic staff viewed sequential images of each specimen and indicated when they first perceived mould. Average human detection times ranged from 145.3 to 185 hours across specimens, with a mean of 164.0 hours, and chi-squared tests confirmed participants were responding to consistent visual cues rather than guessing. The contrast data were then fitted with a three-parameter Gompertz growth model, a sigmoidal function commonly used for microbial growth kinetics, from which a lag time preceding rapid colonisation was estimated for each specimen.

The agreement between the two approaches was close. Fitted lag times ranged from 138.9 to 185.4 hours, and absolute differences from human detection times spanned just 0.3 to 24.7 hours. On average, human detection occurred 6.5 hours after the estimated lag point, suggesting the image method registers structural changes at or just before they become perceptible. A Bland-Altman analysis found a mean difference of 3.45 hours between the two measures with 95 percent limits of agreement from -20.31 to 27.20 hours, and all specimens fell within those limits. The largest discrepancy, specimen 4 at 24.7 hours, hints that human observers can miss early mould development when surface changes are subtle, precisely the scenario in which an objective metric adds most value.

The authors are careful to frame the conditions correctly. The wetting step and elevated humidity represent deliberately severe, accelerated scenarios rather than typical indoor exposure, and the rapid, extensive growth within ten days does not mean mycelium composites are unsafe in general. Rather, it indicates that leaks in humid environments are a high-risk scenario for these materials and underscores the absence of mandatory standardised mould testing for most new building materials, a gap the researchers describe as itself a safety concern. The literature they cite is consistent with their findings: mineral wool and glass wool resisted colonisation even after soaking in one cited study, expanded polystyrene showed less than 10 percent coverage after 90 weeks at 96 to 97 percent relative humidity in another, while hemp and straw insulations developed mould within days to months at elevated humidity.

The image-based method itself is deliberately simple. It requires no powerful or expensive imaging equipment, no thresholding or intensity normalisation when illumination is held constant, and no machine-learning training data, distinguishing it from automated mould-classification approaches that demand large datasets and computational resources. The authors suggest it could be calibrated for different materials by normalising contrast to each substrate type, adapted for in situ monitoring after water-ingress events using flash photography for consistent lighting, and integrated into routine quality control and certification workflows. They also note open questions for future work, including whether the mycelium itself, which contains chitin, serves as a mould nutrient, how hygroscopic water uptake influences risk, and how sterilisation and drying regimes shape the residual microbiome of the finished material. Longer-term biological stability at more moderate humidity remains to be assessed, and real-world field tests are needed to establish safe-use parameters such as critical relative humidity thresholds. As bio-based insulation moves closer to mainstream construction, the study argues, objective monitoring tools like this one will be essential to ensuring these low-carbon materials perform durably in the built environment.

Subject of Research: Mould susceptibility of mycelium-based composite insulation and an image-based method for tracking its onset

Article Title: Image-based evaluation of mould susceptibility in mycelium-derived biomaterials for sustainable construction

Article References: Wildman, J., Cascione, V., Henk, D., & Shea, A. (2026). Image-based evaluation of mould susceptibility in mycelium-derived biomaterials for sustainable construction. Biotechnology for Sustainable Materials, 3(1), Article 1. https://doi.org/10.1186/s44316-025-00051-0

Image Credits: AI Generated

DOI: 10.1186/s44316-025-00051-0

Keywords: mycelium-based composites, mould susceptibility, bio-based insulation, GLCM contrast, image analysis, hemp-shiv, Ganoderma curtisii, building materials, durability testing, humidity, time-lapse imaging, sustainable construction